Title: PicoRadio Protocols, Architectures, and Platforms Picoradio, protocol, intercom, and reconfigurable
1PicoRadioProtocols, Architectures, and
PlatformsPicoradio, protocol, intercom, and
reconfigurable working groupsBWRC
RetreatBerkeley - June 1999
2PicoRadio Goals
- Develop meso-scale radios for ubiquitous
wireless data acquisition that minimize
power/energy dissipation - Minimize energy (lt10 pJ/(correct) bit) for
energy-limited source - Minimize power (lt 1 mW) for power-limited source
(e.g. based on energy scavenging) - By using the following strategies
- self-configuring networks
- fluid trade-off between communication and
computation - aggressive low-energy architectures and circuits
3PicoNode for Sensor Networks
- Single-chip node provides all communication,
geolocation and computation functions, necessary
for an adaptive distributed sensor network - Main Premises
- integration leads to lowest cost, size, and
energy - integration of communication and computation
enables fluid optimization of communication
versus compression, depending upon system
requirements and environment
4Milestones
- Year 1
- Prototype test board to demonstrate distributed
communications/computations - Debugging and monitoring software environment
- Year 2
- Prototype PicoNode (using board-level solution)
- API into wrieless network stack to enable dynamic
variation of transmission parameters and
functions - Year 3
- Integrated PicoNode
- Software environment for remote programming
5Specifications
- Reactive and/or self-triggered bursty
transmissions - Data Sources large number of sensor nodes
- Data Sinks limited set of monitors and/or data
miners - Limited Source Data Rate 100 to 1 Kbit/sec
- System size 1m to 100m, 100 to 1000 nodes
- 5 Ghz band
6Identified Application Areas
- Smart tagging and identification
- Environment monitoring and control
- Security
- Smart periphery
- Instantaneous networking
- class-room, audience participation
7Energy/Power is the central focus!
- - foremost low-energy wireless system design
- adaptive self-configuring wireless
systemsDynamic trade-off between computation and
communication is the key to system-level energy
optimization - energy minimization through all the abstraction
layersapplication, network, media access,
physical - - implementation methodologies to enable the
above - - radio architectures - what needs to be adaptive
or what is fixed? - - circuit implementation - how the get the best
buck for the pJ - - approaches to energy-scavenging
- - design methodology
- - self-configuring and adaptive protocols,
communication channel design - - automating the design generation process
8Research Thrusts
- Main focus on network and protocol processing,
low-energy design, and implementation platforms - Themes
- Source coding for sensor networks
- Wireless networks , protocols and associated
design methodology - Embedded software and RTOS to support dynamic
system reconfiguration - Prototyping
- Low-energy implementation platforms and fabrics
- Alternative ways of communication and transmission
9Source Coding for Energy Reduction
- Correlation between data sources results in
significant redundancy - Coordination between sources results in extra BW
- If statistical correlation structure is known,
compression without collaboration is possible
(theoretical result) - We propose practical constructive framework
without collaboration proposed that achieves
performance of distributed source-coding with
collaboration DISCUS Distributed Source Coding
Using Syndromes (Ramachandran)
10Piconode Foundations
1 megawattfor 100Kbps!
Assumes R-4 loss due to ground wave (_at_ 1 GHz)
90dBm
90dBm
100 Kbps
50dBm
50dBm
10dBm
Transmit Power
10dBm
Transceiver Power
-30dBm
-30dBm
-70dBm
-70dBm
1m
10m
100m
1Km
10Km
Distance
11Energy Minimization at the Protocol Level
- Dominant factor in energy equation determined by
data rate and distance requirements - for cells smaller than 10 m transmitting 1
Kbits/sec communication energy can be ignored.
Energy-efficient computation is key. - This is not the case if the distance is increased
to 100 m (size of a home). Minimization of
communication energy becomes a prime driver.
Partitioning of the link and the use of repeaters
is beneficial (similar to interconnect on chips -
but much more outspoken) - Finding the right optimum is even-harder in
self-configuring systems - precise location and communication requirements
of subscribers not known in advance and vary over
time - no (or little) background infrastructure or
coordination
12Multi-Hop Networks
- Protocols for ad hoc wireless networks
- Destination-Sequenced Distance Vector
- Dynamic Source Routing
- Temporally-Ordered Routing Algorithm
- Ad Hoc On-Demand Distance Vector
- Wireless LAN
13Current focus Simulation and Modeling
Environment
- Evaluation of network modeling toolsNS,
UCB/LBNL/VINT Network Simulator, Optnet, Glomo - Investigation of routing protocols that minimize
Energy Cost Function - Definition of Energy Metric
- Cost F(cost per hop, of hops, ???
14Mac-Level Energy Minimization
- synchronous access control probably most
energy-efficient, but hard to implement in
environment with large number of transceivers
incurs protocol communication overhead - non-coordinated asynchronous access incurs some
inefficiency due to access conflicts, but has
advantage of reduced coordination overhead and
potentially can evolve to energy-efficient
optimum simpler and more robust
15Protocol Design Methodologies
- Specification
- formally describing what the protocol is supposed
to do - Verification
- is the protocol logically consistent?
- Performance Estimation
- is the protocol efficient?
- Implementation
- building a system that implements the
specification
Complicated by adoption of dynamically adapting
protocols
16Refinement-based Protocol Design
- Synthesize protocols
- functionally correct by construction
- satisfying performance requirements
- low power
- dynamically adapting to changing conditions
- reusable
- . by exploring the design space and evaluating
tradeoffs like - energy consumption vs error rate
- bandwidth vs. delay
- computation vs. communication
17Refinement-based Design Flow
System Spec
Input language (ECL,SDL)
Co-simulation Formal Verification
Model of Computation (CFSMs)
Refinement
Stateflow/Simulink
Implementation Hw (VHDL) / Sw (C)
18PicoNode Prototyping Digital Intercom
Basestation
Implementation Platform
Embedded StrongARM Processor
Programmable Logic
1.6 Mbit/sec FH Radio (Proxim)
Mobiles
Up to 20 users per cell _at_ 64 kbit/sec per
link TDMA selected as MAC protocol
Towards single chip
Exercises the complete design flow from
high-level specification
19A prototyping environmentDigital Intercom
- Intended as wireless testbed and prototyping
environment for picoradio - Initial implementation based on Infopad chassis
and off-the-shelf hardware - Designed to allow for interchangeable radio
modules (DECT, Frequency hopping, CDMA) - Software support includes RTOS and wireless
protocol stack
20Architecture Level
- Energy-efficiency dictates custom implementation
of often recurring functions - Adaptivity and configurability requires
programmability for protocol and communication
processing - reconfigurable platforms enable trade-off between
efficiency and flexibility - Pleiades established reconfigurable approach for
data-flow style computation - novel approaches being examined in the protocol
and control space
21Wireless Algorithms Beat Moores Law
Log Complexity
1982 1992 2002
2012
Time
22PicoNode for Sensor Networks
Heterogeneous Implementation Architecture
allows for Trade-off between Flexibility and
Efficiency
23Architectural Choices
Flexibility
1/Efficiency
24Multi-granularity Reconfigurable Architecture
The Berkeley Pleiades Architecture
Configuration Bus
- Computational kernels are spawned to satellite
processors - Control processor supports RTOS and
reconfiguration - Order(s) of magnitude energy-reduction over
traditional programmable architectures
25Maia Reconfigurable Baseband Processor for
Wireless
- 0.25um tech 4.5mm x 6mm
- 1.2 Million transistors
- 40 MHz at 1V
- 1 mW VCELP voice coder
- Hardware
- 1 ARM-8
- 8 SRAMs 8 AGPs
- 2 MACs
- 2 ALUs
- 2 In-Ports and 2 Out-Ports
- 14x8 FPGA
26Low-Energy Embedded FPGA
- Test chip
- 8x8 CLB array
- 5 in - 3 out CLB
- 3-level interconnect hierarchy
- 4 mm2 in 0.25 mm ST CMOS
- 0.8 and 1.5 V supply
- Simulation Results
- 125 MHz Toggle Frequency
- 50 MHz 8-bit adder
- energy 70 times lower than comparable Xilinx
- Parameterized module generator available
27Architecture Comparison
LMS Correlator at 1.67 MSymbols Data
Rate Complexity 300 Mmult/sec and 357 Macc/sec
16 Mmacs/mW!
Note TMS implementation requires 36 parallel
processors to meet data rate - validity
questionable
28Software Methodology Flow
29Implementation Fabrics for Protocols
Intercom TDMA MAC
30Intercom TDMA MACImplementation alternatives
- ASIC 1V, 0.25 mm CMOS process
- FPGA 1.5 V 0.25 mm CMOS low-energy FPGA
- ARM8 1 V 25 MHz processor n 13,000
- Ratio 1 - 8 - gtgt 400
31Low-energy Circuit Design
- Dropping voltage down to minimum possible levels
(between 100 and 500 mV) - Study potential of sub-threshold operation, and
activity-based threshold control - Study impact of burst-mode operation on circuit
design e.g. potential of dynamic voltage
scaling, current-mode logic, and adiabatic design - Study impact of performance variations on circuit
architecture self-timing or mesochronous designs
32Performance/energy trade-off
- Possible control parameters supply voltage,
threshold voltage, and current - Current-control attractive because of minimum
overhead in circuitry or technology
33CMOS versus CML
Logic Depth 4
34Dependence upon process variations
35The Electro-Mechanical Aspects
- How to build meso-scale (millimeter-size)
radios? - enabling energy-scavenging
- how to deal with the antenna issue?
- packaging
- Prototyping experiments in ME 221 (F99)
36Means of energy-scavenging
- Motion (I.e. wrist watch, IDs, pens)
- Light (sensors)
- Pressure
- Recycle energy from cheap down-link to expensive
up-link - Of course, batteries are always an option if one
can keep the energy dissipation of the component
ignorable!
37Summary
- PicoRadio Driver in early stages of conception
- System specification and killer apps are being
defined - Some substantial progress on technology base